Role of cholesterol in SNARE-mediated trafficking on intracellular membranes

Cholesterol homeostasis is essential for the functional integrity of the cell and requires a tight regulation of cholesterol levels inside and outside of cells. Therefore, cholesterol-sensing regulatory circuits communicate any alterations in cholesterol levels in order to control sterol synthesis, uptake and transport (Ikonen, 2008; Maxfield and van Meer, 2010).

Cholesterol is considered to be indispensable for cell migration. A substantial body of literature suggests that specialized cholesterol-rich microdomains and cholesterol-sensitive transport pathways are required for the delivery of integrins and extracellular matrix (ECM) to and from the cell surface. Yet, despite the now well-recognized increased cholesterol demand in cancer cells, it remains unclear how intracellular cholesterol transport routes are linked with the molecular machinery that controls cell migration.

Here, we will first summarize our current knowledge of cholesterol homeostasis, in particular the cellular distribution of low-density lipoprotein (LDL), which is linked to events that are highly relevant for cell migration. We will then focus on how cholesterol transport along intracellular routes can modulate the localization of several NSF attachment protein (SNAP) receptor (SNARE) proteins, major players in membrane transport to the cell surface along endocytic and exocytic pathways. Finally we will discuss how localization of cholesterol-sensitive SNAREs, in particular within Golgi–endosomal boundaries, determines the efficacy of delivery of cell adhesion receptors and ECM, which are key factors in cancer cell migration, to the cell surface.

Cholesterol is an amphipathic molecule with a hydrophilic hydroxyl group and a hydrophobic section organized into four rigid rings followed by a short branched hydrocarbon tail. Within the membrane, the hydroxyl group of cholesterol interacts with the polar head of phospholipids and sphingolipids, whereas the hydrophobic part remains immersed in the membrane alongside the hydrocarbon chains of the surrounding lipids. Cholesterol is not uniformly distributed in the membrane, and besides its interaction with membrane lipids, interaction with proteins also have an important role in its distribution (Epand, 2006; Ikonen and Jansen, 2008). Consequently, there has been a considerable interest on deciphering the molecular mechanisms that allow these interactions. Indeed, several posttranslational modifications appear to enable proteins to preferentially associate with cholesterol-rich domains, including their myristoylation, palmitoylation or glycosylphosphatidylinositol linkage. However, for cholesterol to interact with transmembrane domains of proteins, one would expect to find specific molecular features at the lipid–water interface region of proteins such as defined amino acid sequences or motifs. Indeed, several short cholesterol-binding motifs that are located in the lipid–water interface region favor interactions with cholesterol (Box 1). In some cases, the precise positioning of a cholesterol-binding motif within the lipid–water interface segment might be influenced by the ‘snorkelling’ effect seen for lysine and arginine residues with long buried hydrophobic side-chains (Chamberlain et al., 2004; Strandberg and Killian, 2003). However, how such interactions contribute to maintaining a heterogeneous distribution of cholesterol within the cell and to cholesterol movement between cellular membranes is not fully understood.

Several reviews have discussed the intracellular distribution of cholesterol and cholesterol-rich membrane domains, pathways of intracellular cholesterol movement and methods for studying sterol transport and distribution (Ikonen, 2008; Maxfield and van Meer, 2010; Mesmin and Maxfield, 2009). In general, cells acquire cholesterol through receptor-mediated endocytosis of LDL or by de novo synthesis in the endoplasmic reticulum (ER) (Simons and Ikonen, 2000). After LDL internalization, esterified LDL-cholesterol traffics through the endocytic compartment to be delivered to late endosomes and lysosomes, where it is hydrolyzed by lysosomal acid lipase to free the cholesterol molecule. The majority of late-endosome-derived cholesterol is then delivered to the plasma membrane, while some of it is transported to the ER to enable feedback control or undergoes esterification for storage in lipid droplets in the form of cholesteryl esters. The routes and mechanisms by which late endosome-originating cholesterol reaches the plasma membrane are not clear, but as shown by our laboratory (Cubells et al., 2007; Reverter et al., 2014) and others (Kanerva et al., 2013; Urano et al., 2008), might involve transport through the ER, the trans-Golgi network (TGN) and recycling endosomes (Fig. 1A). As outlined below, although recycling endosomes and the TGN contain much less cholesterol than the plasma membrane, moderate changes in the levels of cholesterol that are transported through these compartments appear to have drastic effects on cellular behavior that is relevant in the context of migration and invasion.

The physiological importance of LDL uptake and lysosomal processing is underscored by genetic defects that cause familial hypercholesterolemia, as well as Niemann Pick type C (NPC) disease, a lethal condition that is characterized by intracellular accumulation of unesterified cholesterol in late endosomes and lysosomes. Niemann Pick C1 protein (NPC1) and Niemann Pick type C2 protein homolog (NPC2) bind to late-endosome-associated cholesterol and facilitate its export from late endosomes (Ikonen, 2006; Vance and Karten, 2014). Mutations in the genes encoding NPC1 or NPC2 inhibit the egress of cholesterol from late endosomes (Klein et al., 2006) and reduce its delivery to the Golgi, plasma membrane and recycling endosomes (Cubells et al., 2007; Kanerva et al., 2013; Reverter et al., 2014; Urano et al., 2008). Together, this triggers a dysfunction of membrane trafficking, which is associated with cardiovascular, neurological and lysosomal storage diseases (Cortes et al., 2013; De Matteis and Luini, 2011; Ikonen, 2006; Maxfield and Tabas, 2005). (Pre-)lysosomal cholesterol export most likely involves the transfer of cholesterol from luminal NPC2 to membrane-associated NPC1, followed by insertion of the aliphatic side chain of cholesterol into the (pre-)lysosomal membrane (Infante et al., 2008; Wang et al., 2010). Although subsequent cholesterol transfer to other sites is not fully understood, non-vesicular transport mediated by cytosolic cholesterol-binding proteins and vesicular pathways exist (Fig. 1B,C). In addition, membrane contact sites between late endosome or lysosomes and other compartments, such as the ER, appear to enable cholesterol transfer (Alpy et al., 2013; Chang et al., 2006; Ikonen, 2008; Rowland et al., 2014; van der Kant and Neefjes, 2014).

It should be noted that in some cell types, in particular hepatocytes and steroidogenic cells, uptake of cholesterol from high-density lipoprotein (HDL) can significantly contribute to cholesterol homeostasis. This is mediated by the scavenger receptor class B member 1 (SRB1, also known as SR-BI), an integral membrane protein that is highly expressed in liver and steroidogenic tissues. SRB1 facilitates the incorporation of HDL-derived cholesteryl esters through a process called selective uptake (Leiva et al., 2011). This is not associated with HDL particle internalization through endosomal or lysosomal pathways, but involves the binding of HDL to SRB1 at the cell surface, followed by passive diffusion of HDL-derived cholesteryl esters into the plasma membrane. Internalized cholesteryl esters are then rapidly hydrolyzed. Consequently, HDL-derived cholesterol quickly enters other sites, such as the ER and recycling endosomes, with kinetics that are much faster than for trafficking of LDL-derived cholesterol (Heeren et al., 2006; Leiva et al., 2011). Therefore, cholesterol generated by this means can rapidly affect the distribution of cholesterol-sensing SNAREs in these compartments (see below).

In a recent comprehensive survey using photoreactive sterol probes in combination with quantitative mass spectrometry, Hulce and co-workers identified over 250 cholesterol-binding proteins, including several SNARE proteins (Hulce et al., 2013).

More than 60 SNAREs have been identified from yeast and mammals, and they are known to be crucial components of protein complexes that drive membrane fusion in secretory and endocytic pathways (Jahn and Scheller, 2006). Site-specific interaction and pairing of SNAREs on target membranes (t-SNAREs) with SNAREs on vesicles (v-SNAREs) facilitates tethering, docking and fusion of distinct vesicle-mediated transport events. This complex and multifactorial process determines the efficiency and speed of delivery of secretory vesicles. Although SNAREs were originally classified as v- and t-SNAREs according to their localization, they can be structurally distinguished as Q or R types (Hong, 2005; Hong and Lev, 2014; Südhof and Rothman, 2009) and further segregated into four subfamilies. These are the syntaxin subfamily (Qa), the S25N (Qb) and S25C (Qc) subfamilies and members of the VAMP subfamily, which are all R-type SNAREs. All syntaxins (except STX11) and VAMPs are integral membrane proteins and contain a transmembrane domain. SNAP23, SNAP25 and SNAP29 are all members of both the Qb and Qc subfamilies and contain two tandem SNARE motifs, but lack a transmembrane.

Strikingly, from the 38 SNAREs found in humans (Hong and Lev, 2014), 11 have been shown to interact with cholesterol (Fig. 2A). However, little structural information on cholesterol–protein complexes is available, and despite the fact that crystal structures of single or complexed SNAREs exist, information regarding their interaction with cholesterol is scarce. One exception is the relatively well-characterized interaction between VAMP2 and cholesterol. Here, site-directed spin labeling, continuous wave and pulsed electron paramagnetic resonance spectroscopy have shown that a significant conformational change takes place in the transmembrane domain of VAMP2 upon cholesterol binding, which is thought to support lipid mixing and eventually membrane fusion (Tong et al., 2009).

However, only one of the cholesterol-binding SNAREs is found at the plasma membrane, the t-SNARE STX4 and its partner SNAP23, which binds to cholesterol through palmitoylated cysteine residues located in the linker region between the two SNARE domains (Salaün et al., 2005) (Table 1). Furthermore, several of the SNAREs that are capable of binding to cholesterol are not known to function in a cholesterol-sensitive manner, but represent SNAREs that are predominantly found on endomembranes, often shuttling between locations, such as late endosomes and recycling endosomes, as well as the Golgi. Thus, the potential interaction between cholesterol and these SNAREs on endomembranes could point to an exciting new molecular pathway that couples local cholesterol levels in endosomal membranes and the Golgi with dynamic changes in SNARE location and thereby their function in cell dynamics and behavior.

In line with 70–80% of cellular cholesterol being found at the plasma membrane, several studies have identified a crucial role for cholesterol in inducing and stabilizing the clustering of SNAP23 and STX4, which is important for membrane fusion (Chamberlain et al., 2001; Lang, 2007; Predescu et al., 2005; Puri and Roche, 2006). SNAP23 binds to STX4 in vivo (St-Denis et al., 1999). Accordingly, together these two SNAREs are involved in the fusion of transferrin-containing recycling vesicles with the basolateral membrane (Leung et al., 1998), as well as surface delivery of newly synthesized proteins from the Golgi to the apical membrane in Madin–Darby canine kidney (MDCK) cells (Lafont et al., 1999). Although we and others confirmed that these SNAREs, similar to STX1 or SNAP25, concentrate within clusters at the plasma membrane (Reverter et al., 2011; Sieber et al., 2006; Sieber et al., 2007), the underlying causes that promote STX4–SNAP23 clustering are not fully understood.

Given the high cholesterol content at the plasma membrane and its membrane-condensing effect, which increases membrane order and rigidity (Lingwood and Simons, 2010), the physicochemical properties of cholesterol probably play a major role in the formation of SNARE microdomains at the plasma membrane (van den Bogaart, 2013). In addition, factors that promote SNARE clustering at the plasma membrane include their partitioning into cholesterol-enriched membrane rafts, competition with cholesterol for solvation by bulk lipids, and electrostatic protein–lipid interactions. Furthermore, the clustering of plasma-membrane-associated proteins, including the SNAREs listed above, might be promoted by protein–protein interactions (Zilly et al., 2011), homotypic interactions between the SNARE motifs (Sieber et al., 2006), heterotypic protein–protein interactions (Yang et al., 2006) and anchoring to the cortical cytoskeleton (Low et al., 2006; Torregrosa-Hetland et al., 2011).

Nevertheless, the extent to which cholesterol-mediated plasma membrane clustering of SNAREs contributes to their overall distribution remains unclear. The reported effects of cholesterol on SNARE clustering are predominantly based on extraction of cholesterol from cells under non-physiological conditions with methyl-β-cyclodextrin, which is known to cause a partial to complete disintegration of plasma membrane clusters that are enriched in STX1, SNAP25, STX4 and/or SNAP23 (Low et al., 2006; Predescu et al., 2005; Reverter et al., 2011). Moreover, within a single cell type, cholesterol extraction can completely disrupt plasma membrane clusters of a particular SNARE, whereas other plasma-membrane-associated SNARE clusters remain unaffected, suggesting that the extent of cholesterol-mediated clustering differs among SNAREs and cell types. The underlying mechanisms have yet to be resolved, but certain cholesterol pools located in specific microdomains might be more accessible and susceptible to depletion by experimental protocols, which could lead to a sequential disruption of membrane microdomains with variations in their SNARE and cholesterol content.

The late endosomal compartment is not only central to cholesterol homeostasis (see above), but recent data are beginning to unravel how late endosomes and lysosomes influence the dynamics of adhesions between cells and the ECM and movement of cells in both two- and three-dimensional microenvironments (Rainero and Norman, 2013). For instance, the endosomal sorting complexes required for transport (ESCRT), which regulate delivery of cargo and membrane transport along the endocytic pathway, control the transit of the tyrosine kinase Src from late endosomes to focal adhesions, adhesion structures required for cell–ECM interactions, at the plasma membrane. At the cell surface, Src is involved in mediating the disassembly and turnover of cell–ECM interactions, which are required for cell migration (Tu et al., 2010). Particularly relevant for cancer metastasis, oncogenic v-Src kinase is located in perinuclear structures that resemble late endosomes, and it translocates to the plasma membrane to mediate cell transformation (Fincham et al., 1996). Along these lines, in pancreatic and ovarian cancers, Rab25 together with the chloride intracellular channel 3 (CLIC3) drives invasiveness by recycling of α5β1 integrin from late endosomes to the plasma membrane (Dozynkiewicz et al., 2012).

The protein machinery that controls the fusion of late endosome and lysosomal compartments with the plasma membrane is now relatively well documented and involves syntaxin 7 (STX7), STX8, vesicle-associated membrane protein 7 (VAMP7) and VAMP8, which are all members of the SNARE protein family (Kent et al., 2012; Luzio et al., 2010; Luzio et al., 2014; Pryor et al., 2004). Chavrier and co-workers have established that VAMP7 and the exocyst complex deliver a late endosomal pool of membrane type 1 metalloproteinase (MT1-MMP, also known as MMP14) to the cell surface for ECM proteolysis, which is crucial for invasive migration (Steffen et al., 2008). Earlier steps in this pathway might involve CLIC3 (Macpherson et al., 2014). Later trafficking events probably require the actin cytoskeleton, as following its transit from Rab7-positive endosomes to the plasma membrane, MT1-MMP is retained in invasive protrusions by direct tethering to F-actin (Yu et al., 2012).

In addition, besides these SNARE-dependent trafficking events that emanate from late endosome and drive cell migration, distribution of LDL-derived cholesterol from lysosomes and late endosome to other cellular sites has now been recognized to also trigger trafficking events that promote cell migration. Recently, Ikonen and co-workers identified the delivery of LDL-derived cholesterol from late endosomes to focal adhesions (Kanerva et al., 2013). As outlined above, cholesterol is essential for the functioning of SNAREs at the plasma membrane and one could envisage that the compromised motility and function of late endosome and lysosomes in lysosomal storage diseases such as NPC – which is caused by accumulation of cholesterol and/or sphingolipids (reviewed in detail in Fraldi et al., 2010; Ikonen and Hölttä-Vuori, 2004; Platt et al., 2012) – interferes with the functioning of those SNAREs that are localized in the late endosome compartment and are relevant for cell migration (Ganley et al., 2008).

By using NPC1-mutant-like cell models that accumulate cholesterol in the late endosomes, by pharmacological means (treatment with U18666A) or annexin A6 overexpression, which inhibits NPC1 activity possibly through direct protein–protein interaction (Cubells et al., 2007), we initially showed that blocking egress of cholesterol from late endosomes interfered with cholesterol-dependent caveolin-1 transport from the Golgi to the cell surface, thereby reducing the number of caveolae (Cubells et al., 2007). We speculated this to reflect earlier data from in vitro studies showing that precisely balanced cholesterol levels are essential to drive vesicle formation from the Golgi (Stüven et al., 2003). Although cholesterol depletion inhibits the formation of secretory vesicles from the TGN (Wang et al., 2000), elevated cellular cholesterol stimulates vesicle formation and the dispersal of Golgi-derived vesicles (Grimmer et al., 2005). Efforts to identify the cholesterol-sensitive factor that is involved in this process found that cytoplasmic phospholipase A2 (cPLA₂) translocates to the Golgi upon elevation of cellular cholesterol (Grimmer et al., 2005). Based on these data and the fact that cPLA₂ converts phospholipids into lysophospholipids that have an inverted-cone shape, the authors hypothesized that cPLA₂ contributes to Golgi membrane deformations that are required for the formation of secretory vesicles, as pharmacological inhibition of cPLA₂ has been shown to inhibit cholesterol-induced vesiculation at the Golgi (Grimmer et al., 2005).

In further support of this hypothesis, our follow-up studies revealed that a decrease in the availability of cholesterol in the Golgi of NPC1-mutant-like cells upon using pharmacological treatment with U18666A or annexin A6 overexpression perturbed the translocation to and activity of cPLA₂ at the Golgi (Cubells et al., 2008). We speculated that inhibition of cholesterol egress from late endosomes leads to a depletion of Golgi cholesterol, therefore disrupting cholesterol-dependent cPLA₂ recruitment and vesiculation events at the Golgi (Cubells et al., 2008). This would interfere with cholesterol delivery to the cell surface along exocytic pathways that originate at the Golgi. As the plasma membrane undergoes rapid and continuous turnover, the lack of cholesterol replenishing the cell surface pool would be followed by the disintegration of STX4 and SNAP23 clusters that form at cholesterol-enriched plasma membrane microdomains. Consistent with this model, pharmacological inhibitors of cPLA₂ strongly reduced the localization of SNAP23 and STX4 at the plasma membrane (Reverter et al., 2011). Moreover, depletion of the cPLA₂ product, arachidonic acid, has been shown to inhibit STX4 from forming SNARE complexes (Connell et al., 2007). Arachidonic acid induces the so-called ‘syntaxin-opened’ conformation that is required for SNARE assembly (Connell et al., 2007; Rickman and Davletov, 2005), further supporting the involvement of cPLA₂ in the modulation of syntaxins that participate in SNARE trafficking and assembly.

Most strikingly, we found that although clusters of SNAP23 and STX4 at the plasma membrane disintegrated in NPC1 mutant-like cells, both SNAP23 and STX4 accumulated together with caveolin-1 in Golgi membranes, which as we shown, is associated with an increased formation of SNAP23–STX4-containing t-SNARE complexes in the Golgi (Reverter et al., 2011). Consequently, the accumulation and mislocalization of SNAP23–STX4 complexes in the Golgi of these NPC1-mutant-like cells prevented SNAP23–STX4 from promoting the secretion of fibronectin, an ECM component that binds to cell adhesion receptors, at the plasma membrane. Mislocalization and loss of SNAP23–STX4 function upon cholesterol depletion might also occur in other STX4–SNAP23-dependent events that are relevant for cancer cell migration, for instance the association of STX4–SNAP23 with VAMP7, which is required to deliver MT1-MMP to invadopodia for ECM degradation in invasive cancers (Williams et al., 2014).

Thus, the depletion of cholesterol from the Golgi and plasma membrane owing to its accumulation in late endosomes not only interferes with the localization and organization of some SNAREs at the plasma membrane, but also impacts on the dynamics of SNARE assembly and disassembly on intracellular membranes, such as at the Golgi and transport pathways originating from the Golgi (Reverter et al., 2011; Reverter et al., 2014).

Based on the studies discussed above, it is tempting to speculate that the mislocalization of SNAP23–STX4 upon NPC1 inhibition is only an example of the broader impact a cellular imbalance of cholesterol has on endomembrane SNAREs, including those involved in post-Golgi exocytic pathways through the TGN and recycling endosome compartments and those relevant for cell migration.

Integrins are cell adhesion receptors; they are composed of α and β subunits that bind to the ECM and enable cells to migrate. To ensure forward movement, integrins are constantly internalized and recycled for their redistribution at the leading edge (Caswell and Norman, 2008; Jones et al., 2006; Pellinen and Ivaska, 2006). Both cholesterol and several SNAREs have been implicated in the molecular machinery that is responsible for the endocytic and exocytic trafficking of integrins. As such, the cholesterol content of the plasma membrane controls signaling events that are mediated by αVβ3 integrins (Green et al., 1999), including focal adhesion kinase (FAK) and mitogen-activated protein kinase (MAPK) signaling pathways, and cell adhesion and migration on fibronectin-coated substrates (Ramprasad et al., 2007). In this context, the ability of SNAREs to bind cholesterol in specific cellular sites could contribute to the establishment of functional links between cholesterol and integrin localization and function. These cholesterol-binding SNAREs include VAMP2, VAMP3 (also known as cellubrevin), STX3, STX4 and SNAP23, which all participate in the recycling of β1 integrin, and VAMP3 and STX6, which determine the levels of cell-surface-associated α5β1 integrin and of FAK (Hulce et al., 2013; Riggs et al., 2012; Tiwari et al., 2011) (see also Fig. 2A and Table 1).

Among the aforementioned v-SNAREs, VAMP3 contains a cholesterol-interacting R-W-L motif (Fig. 2B). VAMP3 is highly similar to VAMP2 and has been shown to bind to cholesterol in exocytic pathways (Table 1). Besides its similarity to VAMP2, VAMP3 also displays a significant sequence similarity to VAMP4, which might contribute to the ability of both VAMPs to bind the t-SNAREs STX6, STX16 and VIT1A, which function at the trafficking interface between the TGN, recycling endosome and the plasma membrane. However, VAMP4 does not appear to bind cholesterol, but contains an E/DxxxLL motif that allows interaction with adaptor protein-1 in the TGN (Peden et al., 2001). These differential binding properties might enable VAMP3 and VAMP4 to participate in different trafficking pathways that intersect at the TGN, where they interact with their common t-SNARE partner STX6.

As discussed above, both VAMP3 and VAMP4 interact with the t-SNARE STX6, which, at steady state, is predominantly localized at the TGN where it binds to resident molecules through its SNARE motif (Bock et al., 1997; Wendler and Tooze, 2001) and Habc domain (Abascal-Palacios et al., 2013; Pérez-Victoria et al., 2010). Besides VAMP4, this includes the Golgi SNARE protein GS32 (also known as SNAP29) and the mammalian Golgi Sec1p-like protein VPS45, both of which are known to be involved in TGN trafficking (Watson and Pessin, 2000). STX6 participates in various membrane fusion events with different SNARE complexes. Upon completion of membrane fusion, a tyrosine-based sorting motif (YGRL, position 140–143) located between the N-terminal domain and the SNARE motif is activated, which results in its re-sorting to the TGN (Jahn and Scheller, 2006; Jung et al., 2012). Importantly, STX6 binds to cholesterol (Hulce et al., 2013) and has been previously implicated in cholesterol transport; it contributes to the delivery of late-endosome-derived cholesterol to the ER via the TGN (Urano et al., 2008) and of lipids and proteins that are required for caveolae endocytosis to the plasma membrane (Choudhury et al., 2006). This work has shown that inhibition of STX6 leads to a reduction of caveolin-1 and caveolae at the cell surface, lending further support for a model by which STX6 regulates secretory pathways in a cholesterol-sensitive manner. This ability of STX6 to regulate the transport of cholesterol-rich domains to and from the cell surface could determine the cell surface levels of α5β1 integrin and FAK, and so modulate focal adhesion sites and directional migration on fibronectin-coated substrates (Tiwari et al., 2011).

Taken together, cholesterol is linked to the localization and function of STX6 and other SNAREs that are predominantly located at the TGN, the TGN–recycling-endosome boundaries and recycling endosomes. Using an array of NPC1 mutant models, we were able to provide novel insights into how cholesterol pools at the Golgi–endosomal boundaries regulate cell migration, namely through altering the assembly and disassembly of complexes between STX6 and VAMP3 or VAMP4, respectively (Reverter et al., 2014). Mechanistically, we found that a decrease in the cholesterol level at the Golgi perturbed the trafficking between recycling endosomes and TGN and triggered the accumulation of STX6 in VAMP3- and Rab11-containing recycling endosomes. This correlated with a reduced cell surface expression of integrins in NPC1 mutant models and resulted in impaired cell migration and invasion in two- and three-dimensional environments. Support for endomembrane cholesterol levels altering SNARE localization and function has been provided by the fact that prolonged treatment of NPC1 mutant cells with LDL, which causes its overspill into late endosomes, thereby enabling LDL-cholesterol to bypass the NPC1 mutation and to enter the TGN at later time points, induced the targeting of STX6 back to the TGN. By contrast, incubation of wild-type cells with HDL-cholesterol, which rapidly enters recycling endosomes, but not the TGN, within 30 min after binding of HDL to cell surface receptors, results in a relocalization of STX6 from the TGN to recycling endosomes (Reverter et al., 2014). Thus, the rapid and local elevation of cholesterol levels in recycling endosomes of wild-type cells observed after HDL incubation appears to increase the ability of STX6 to interact with VAMP3 in recycling endosomes.

Further support for a function of STX6 as a cholesterol-sensitive endomembrane SNARE comes from its structure; this syntaxin contains an R-W-L motif (position 233–241) (Fig. 2B), as well as a polybasic cluster in the juxtamembrane region that has homology to STX4, STX7 and STX12. This polybasic region is conserved and also found in STX1A, STX1B, STX2, STX3, STX17, STX20 and STX16 (Murray and Tamm, 2011). In contrast, this positively charged region appears to be less prevalent in syntaxins that are required for early steps in membrane trafficking, such as STX18 in the ER or STX5 in the Golgi. Interestingly, the syntaxins with the most positive charges localize to the plasma membrane in microdomains that are characterized by a high degree of negative charge density and are enriched in cholesterol (Murray and Tamm, 2011). Thus, the spatiotemporal localization and function of the syntaxins that have cholesterol-binding motifs and polybasic clusters, such as STX6, STX4, STX7 and STX12, might be highly responsive to fluctuations in cholesterol levels at either the plasma membrane or intracellular compartments.

The drastic changes in STX6 localization we observed in NPC1-mutant cells and upon incubation with lipoproteins suggest that STX6 trafficking between TGN and recycling endosome and its compartment-specific interaction with the v-SNARES VAMP3 and VAMP4 is controlled by its ability to sense cholesterol levels in the TGN and/or recycling endosomes, thereby possibly regulating cell migration through STX6-dependent trafficking of the integrins αVβ3 and α5β1. NPC1-mediated reduction of cholesterol in TGN membranes or selective elevation of recycling endosome-localized cholesterol could promote the translocation of STX6 to these membranes and so increase its ability to interact with VAMP3 (Reverter et al., 2014). In fact, recycling endosomes are the main intracellular cholesterol repository compartments in a number of cell types, including Chinese hamster ovary and non-polarized hepatoma HepG2 cells, fibroblasts and human B lymphocytes (Hao et al., 2002; Hsu and Prekeris, 2010; Maxfield and McGraw, 2004).

Although other alternative pathways might exist, our findings point to a link between cholesterol-sensitive SNAREs and the final steps in integrin recycling relevant for wound healing and cancer cell migration (Fig. 3). Earlier work has implicated cholesterol in the formation of signaling complexes that contain αVβ3, CD47 and G-proteins (Green et al., 1999) and in the control of cell adhesion and migration onto fibronectin. This possibly requires Rab11, which modulates cholesterol transport and homeostasis (Hölttä-Vuori et al., 2002), and facilitates the recycling of β1 integrin (Powelka et al., 2004). Further support for this notion comes from an increased cholesterol requirement for invasion in MDA-MB-231 breast cancer and A431 epidermoid carcinoma cells (Freed-Pastor et al., 2012); both cell models are highly relevant in this context as enhanced integrin recycling therein drives aggressive cancer cell behavior (Muller et al., 2009). It has yet to be demonstrated whether these observations are based on some SNAREs functioning as cholesterol sensors in cancer. Indeed, STX6 overexpression increases cell migration, and its expression is elevated in breast, liver and prostate cancers (Riggs et al., 2012), which all have established links to cholesterol homeostasis.

Furthermore, in breast cancer cells, SNAP23 and STX12, which also bind to cholesterol in the plasma membrane and recycling endosomes, are required for the delivery of Src, epidermal growth factor receptor (EGFR) and β1 integrin to cholesterol-rich invadopodia for cell invasion. Interestingly, β1 integrin affects the interaction between SNAP23 and STX12 and the formation of the SNARE-dependent Src–EGFR–β1-integrin complex (Williams and Coppolino, 2014). In this context, cholesterol-dependent trafficking mechanisms appear to be essential (Caldieri and Buccione, 2010), and cholesterol-binding proteins such as caveolin-1 (Murata et al., 1995) actively participate in the transport of cholesterol, as well as facilitating its interaction with integrins and their activation (Parton and del Pozo, 2013).

As discussed above, cholesterol levels at late endosomes and the Golgi modulate the localization and activity of a subset of cholesterol-sensing SNARE proteins that regulate fibronectin secretion, integrin recycling and cell migration.

The underlying molecular mechanisms await further investigation. Cholesterol levels in the TGN and recycling endosome are much lower than those in the plasma membrane. Together with the drastic experimental procedures (e.g. treatment with methyl-β-cyclodextrin) required to reduce membrane order at the plasma membrane (Reverter et al., 2014), this might indicate that moderate changes in cholesterol levels are insufficient to significantly alter membrane fluidity within the TGN or recycling endosome subcompartments to affect the affinity of SNAREs towards membranes and thus their distribution. Instead, we envisage that cholesterol transport is targeted directly to or into the vicinity of SNARE complexes, possibly by the docking of cholesterol carriers into close proximity or directly onto endomembrane SNAREs. This could enable SNARE–cholesterol interaction to impact on the stability of assembled SNARE complexes within a certain location, providing the opportunity for these complexes to dissociate and move along transport pathways to find other interaction partners elsewhere. However, it is still unknown whether changes in cholesterol levels affect the three-dimensional structure and conformation of specific SNARE proteins (as postulated for VAMP2) (Tong et al., 2009), or whether cholesterol levels and the lipid environment could interfere with the affinity or capability of certain SNAREs to cluster and/or remain in a particular membrane.

Alternatively, rather than directly targeting endomembrane SNAREs, cholesterol could modulate the activity of Golgi-tethering factors that are required for SNARE assembly. For instance, Golgi-associated retrograde protein (GARP) and conserved oligomeric Golgi (COG) complex regulate the fusion of endosome-derived, retrograde transport carriers to the TGN. This process requires the assembly of the t- and v-SNAREs (STX6, STX16, VIT1A and VAMP4) to facilitate the final steps in the retrograde transport of recycled receptors or integrins, which enter the trans-Golgi for re-sialylation, or transport of Golgi-resident proteins (i.e. TGN38) to TGN–endosome boundaries (Bonifacino and Hierro, 2011). The ability of GARP, COG and other Golgi-tethering complexes to promote or prevent the assembly of the v- and t-SNAREs mentioned above or possibly the pairing of other specific v- and t-SNAREs might depend on their spatiotemporal localization and ability to interact with other cellular SNAREs. Besides the affinity of tethers to their respective SNAREs, the avidity of binding, post-translational modifications or as highlighted here, the amount of cholesterol can influence the assembly or disassembly of SNAREs. The identified role for cholesterol in controlling integrin trafficking through SNAREs could not only be relevant for the ability of cancer cells to move and spread, but is also important for wound healing in diabetes. Undoubtedly, further in vivo and in vitro studies will shed more light on the complex role cholesterol plays in modulating the molecular mechanisms that enable SNAREs to control membrane trafficking and fusion events that are relevant for cell migration in both physiological and pathological contexts.

We would like to thank all members of our laboratories, past and present, for their invaluable contributions and apologize to all those researchers whose work could not be discussed owing to space limitations.